This preview shows pages 1–4. Sign up to view the full content.
Imbens, Lecture Notes 1, ARE213 Spring ’06
1
ARE213
Econometrics
Spring 2006 UC Berkeley Department of Agricultural and Resource Economics
Ordinary Least Squares I:
Estimation, Inference and Predicting Outcomes (W 4.2.14)
Let us review the basics of the linear model. We have
N
units (individuals, Frms, or
other economic agents) drawn randomly from a large population. On each unit we observe
on outcome
Y
i
for unit
i
, and a
K
dimensional column vector of explanatory variables
X
i
=
(
X
i
1
,X
i
2
,...,X
iK
)
±
(where typically the Frst covariate is a constant,
X
i
1
= 1 for all
i
=
1
,...,N
.) We are interested in explaining the distribution of
Y
i
in terms of the explanatory
variables
X
i
using a linear model:
Y
i
=
β
±
X
i
+
ε
i
.
(1)
In this equation
β
is a
K
dimensional column vector. In matrix notation,
Y
=
X
β
+
ε,
where
Y
is an
N
dimensional column vector, and
X
is an
N
×
K
dimensional matrix with
i
th row equal to
X
±
i
. Avoiding vector and matrix notation completely:
Y
i
=
β
1
·
X
i
1
+
...
+
β
K
·
X
iK
+
ε
i
=
K
±
k
=1
β
k
·
X
ik
+
ε
i
.
We assume that the residuals
ε
i
are independent of the covariates or regressors, and normally
distributed with mean zero and variance
σ
2
:
Assumption 1
ε
i

X
i
∼N
(0
,σ
2
)
.
We can weaken this considerably. ±irst, we could relax normality and only assume indepen
dence:
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document Imbens, Lecture Notes 1, ARE213 Spring ’06
2
Assumption 2
ε
i
⊥
X
i
,
combined with the normalization that
E
[
ε
i
] = 0. We can even weaken this assumption
further by requiring only meanindependence
Assumption 3
E
[
ε
i

X
i
]=0
,
or even further, requiring only zero correlation:
Assumption 4
E
[
ε
i
·
X
i
]=0
.
We will also assume that the observations are drawn randomly from some population.
We can also do most of the analysis by assuming that the covariates are Fxed, but this
complicates matters for some results, and it does not help very much. See the discussion on
Fxed versus random covariates in Wooldridge (page 9)
Assumption 5
The pairs
(
X
i
,Y
i
)
are independent draws from some distribution, with the
Frst two moments of
X
i
Fnite.
The (ordinary) least squares estimator for
β
solves
min
β
N
±
i
=1
(
Y
i

β
±
X
i
)
2
.
This leads to
ˆ
β
=(
X
±
X
)

1
(
X
±
Y
)
.
The (exact) distribution of the ols estimator under the normality assumption in Assumption
1is
ˆ
β
∼N
²
β,σ
2
·
(
X
±
X
)

1
³
.
Imbens, Lecture Notes 1, ARE213 Spring ’06
3
Without the normality of the
ε
it is diﬃcult to derive the exact distribution of
ˆ
β
. However,
under the independence Assumption 2 and a second moment condition on
ε
(variance Fnite
and equal to
σ
2
), we can establish asymptotic normality:
√
N
(
ˆ
β

β
)
d
→ N
(
0
,σ
2
·
E
[
XX
±
]

1
)
.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
This is the end of the preview. Sign up
to
access the rest of the document.
This note was uploaded on 08/01/2008 for the course ARE 213 taught by Professor Imbens during the Spring '06 term at University of California, Berkeley.
 Spring '06
 IMBENS

Click to edit the document details